Dhebe Digitals / AI & data / DHEBE-AI-02
AI Evaluation & Data Engineer
Build the evidence behind AI features. Create dependable datasets, evaluation pipelines and quality reports that make improvements and regressions visible.
- Python & SQL
- Dataset quality
- Reproducible evaluations
What you'll work on
- Create versioned evaluation fixtures with explicit provenance, permissions and expected results.
- Build repeatable pipelines for extraction quality, language coverage, regressions and failure analysis.
- Separate training, tuning and evaluation data; report missing evidence rather than hiding it.
What you bring
- Practical Python and SQL skills and experience working with structured datasets.
- Understanding of data leakage, annotation consistency, sampling and quality metrics.
- Ability to document assumptions and communicate where a result does and does not generalize.
First projects
- Create a representative, privacy-safe document evaluation library.
- Add repeatable quality gates before AI-assisted features reach users.
These describe the work area, not a request for free production work as part of an application.
Compensation and engagement
₹80,000 - ₹1,20,000 / month
Ranges are in INR per month, not annual CTC or take-home pay. For full-time employment, the range describes gross monthly base pay; for a paid contract, it describes the monthly fee for the agreed scope. Final compensation stays within the listed range, with hours, deliverables, deductions and engagement terms confirmed in writing before work begins.
Remote within India. This role is available through full-time employment or a paid contract, with the arrangement agreed before starting. Benefits, working hours, equipment and payment schedules depend on the written terms.
YOUR NEXT STEP / DHEBE-AI-02
Apply for this role
Apply to Dhebe Digitals. Your details and resume go to jobs@dhebe.com. We also save your application details in our private assessment database and automatically email a private link for this role at test.dhebe.com. No applicant account, public listing or marketing signup.
The assessment has two parts: 50 challenging role scenarios in 60 minutes, followed by a 60-minute coding task covering core skills for the role. Questions lock when Part 2 starts, and timers continue during interruptions. Questions are scored automatically. Supported coding tasks are checked in our private, isolated grading environment; older exercises receive human review. Recruiters make hiring decisions. We send applicable reminders, submission receipts and results by email.
Before starting, use a desktop or laptop, enable your camera and enter full-screen mode. Work independently and stay in the assessment tab; do not switch tabs, minimise the window or use unapproved assistance. The test asks for your explicit agreement and browser camera permission before it starts. The camera preview stays on your device: no images, video or audio are uploaded or recorded. Limited tab, window-focus, full-screen and camera-status events are saved with timestamps for human review, not automatic failure. Contact us before starting if you need an adjustment or clarification.
After applying, open the private role assessment link sent to your email. Read the instructions and complete the device checks before starting. Completing an assessment does not guarantee selection or employment.